Getting into physics grad school

[Vanadium 50]

Part One: Introduction

There seems to be a lot of discussion and at least some misinformation about the physics grad school admissions process. I thought it might be worthwhile to describe how it works and what departments are looking for. Be aware that every department is unique and does things their own way, so treat this as a general guideline and not absolute fact.

There are two very large differences between undergraduate and graduate admissions. One is that in undergraduate admissions, the student is admitted by a university-wide admissions department. For graduate admissions, while the student may be formally admitted by the university or the school, the decision is made by the physics department. Almost always a faculty committee is set up to do this - around a half-dozen members, representing a range of the department's interests. Sometimes a graduate student is also on this committee, and sometimes there is a member from a closely allied department's faculty - perhaps chemistry or mathematics.

The other difference is that pretty much anyone who wants to go to college can get to go somewhere. That's not true for graduate school. There are around 7000 physics and astronomy graduates per year. About half of them take the GRE, and have presumably at least some interest in graduate school (at least enough to invest the time and expense of the exam), and about half of those ultimately enroll in a graduate program. So every year there are 1000-1500 students with some interest in graduate study in physics who don't get that opportunity.

When the committee meets, they are told by the Department that they have space for X students. That means that they need to offer admission to some larger number of students Y, because some will decline their offer. The ratio X/Y is known as the yield ratio, and departments keep historical records of this, so they know pretty much how many people to admit. They get Z applications, and typically Z >> Y: perhaps 10 or 20 times larger, although of course it varies.

Usually the committee does a first pass through the applications to select ~2Y applicants for a closer look. Usually there is not much arguing at this point - if it's questionable whether a candidate is just a little above or a little below this cut-off, the candidate is probably below the threshold for being offered admission. Also at this point, the candidate's package may or may not have been looked at in detail by all the members of the committee: that can come later. Instead, the committee can divide the applications - for example, if there are 6 members, each may look at 1/3 of the applications in detail, and the other committee members will often just glance at them. Clearly getting on the first pass list is vital. Once there, it's usual for the committee members to look at every application in detail.

(Parts 2-5 will follow)

 

[j93]

The ratio X/Y is known as the yield ratio, and departments keep historical records of this, so they know pretty much how many people to admit. They get Z applications, and typically Z >> Y: perhaps 10 or 20 times larger, although of course it varies.

I hate to nitpick but no school has a 5% acceptance rate
Harvard 12%
Berkeley 16%
from GradSchoolShopper

 

[Dr Transport]

I hate to nitpick but no school has a 5% acceptance rate
Harvard 12%
Berkeley 16%
from GradSchoolShopper

The best resource is the American Institute of Physics, they publish a catalog of grad schools, the faculty listing etc...right down to applications received, accepted, number of degrees granted over the past X years...

If memory serves me correctly, and I could be wrong, but I remember seeing that Rochester accepted single digit percentages (they basically say, if we accept you you will get a PhD) and I'd rank them with Berkely, Stanford, Cornell and some of the other big name schools.

 

[j93]

The best resource is the American Institute of Physics, they publish a catalog of grad schools, the faculty listing etc...right down to applications received, accepted, number of degrees granted over the past X years...

If memory serves me correctly, and I could be wrong, but I remember seeing that Rochester accepted single digit percentages (they basically say, if we accept you you will get a PhD) and I'd rank them with Berkely, Stanford, Cornell and some of the other big name schools.

gradschoolshopper is a site that just links to that aip data. Any data that I have seen that claims a single digit rate is suspect. For example, USC claims they accept 13 out 190 but have 78 grad students. Rochester seems to claim they accept 20 out of 400 but have 114 grad students. They are either flat out lying (cooking the books or they honestly believe accepted students means students who accepted their offers) or have a 100% yield which neither Stanford, MIT , nor Harvard do. Dont take numbers at face value.

 

[AstroChick]

gradschoolshopper is a site that just links to that aip data. Any data that I have seen that claims a single digit rate is suspect. For example, USC claims they accept 13 out 190 but have 78 grad students. Rochester seems to claim they accept 20 out of 400 but have 114 grad students. They are either flat out lying (cooking the books or they honestly believe accepted students means students who accepted their offers) or have a 100% yield which neither Stanford, MIT , nor Harvard do. Dont take numbers at face value.

There are a lot of inaccuracies in the AIP data from gradschoolshopper, from average GPAs and test scores, to the percent of admits. The departments have to report the data themselves, and most departments are too busy to double check their numbers.

 

[AstroChick]

Also, as someone going through the application process this year, thanks for writing this up Vanadium50!

 

[Vanadium 50]

Part 2: Grades:

A physics department invests a lot of effort into educating graduate students. They don't want to admit students that will not complete their degrees, and like it or not, grades are a very strong predictor of how well that person will do.

I don't know what the average GPA is of an admitted student, integrated over all universities, but I would imagine it's around 3.7: the typical student got mostly A's and some B's as an undergraduate. The less competitive one's undergraduate institution is, the higher the expectation of good grades. Below 3.5, a student starts to become uncompetitive very quickly. Below a 3.0 many universities simply will not admit you.

People ask how severe this 3.0 limit is. This varies by school, but it's often taken very seriously. At one university, near the bottom of the rankings of departments, the dean of the college forbids accepting students for graduate admissions with less than a 3.0. Exceptions are granted only by the provost (the senior academic officer of the university). Part of this is because grades once in graduate school are taken seriously: a C is considered failing. When I was a graduate student, if you had any two quarters with either a quarter or cumulative average below 3.0, you were shown the door. The department had no choice in the matter - this was the policy of the college. So they were strongly disinclined to admit students with a history of low grades.

History is an important word here. Committees look at trends and patterns. A history of high grades, backed with strong test scores is the sort of pattern they like. An upward trend in grades is a trend they like. Strong physics grades is a trend they like. Downward trends in grades, they don't like so much. A GPA that offsets low physics grades with higher grades in easy courses is a trend they don't like so much. They look beyond the single number - so all 3.7's are not created equal.

 

[Vanadium 50]

Part 3: Standardized Tests

The graduate equivalent of the ACT or SAT is the Graduate Record Examination or GRE. This comes in two parts, a general test covering verbal reasoning, quantitative reasoning, and critical thinking and analytical writing skills, and a subject test covering what is taught in the typical undergraduate physics curriculum.

The general test is largely irrelevant. Sometimes the college has minimum requirements for the general score, but physics graduates tend not to have any problem with them. Other than that, I have never seen this score make a difference: a student who got in because of a high general GRE or one who was rejected because of a low general GRE.

The key part is the subject test. This is the only way that the committee has to compare across schools: how does a student with a 3.5 at University X compare to one with a 3.6 at University Y? While this test is pretty much universally acknowledged not to be perfect, because it is standardized, it is taken very seriously by committees.

Since only about half of the people who take the GRE go on to graduate school, one needs to score roughly in the top half to be competitive anywhere, and substantially above that if one wants to be competitive at a more selective university.

The other test that's important is the TOEFL, for international applicants. Most departments have had the experience of admitting a bright student from some far-away land, with a great application except for low TOEFL scores. They admitted this student, saying, "look how bright he is - surely he'll pick up English in no time". For whatever reason, this didn't happen, and they ended up with someone with English skills so poor that they couldn't use him as a TA, and whose presentations were very difficult to follow, making his path to a PhD quite rough. Most departments have learned from this experience and are taking increasingly close looks at TOEFL scores. International students should be aware of this.

 

[Vanadium 50]

Part 4: Letters of Recommendation

These are very important. Grades and GREs are just a pile of numbers (correlated ones at that) and don't give as an accurate a view of the candidate as letters do. In many cases, letters are the deciding factor on whether to admit someone or not.

To set the scale, about 1 in 4 students ends up going to graduate school. The average college graduates 10 physics majors per year, so about 2 people per class go. Each student will likely (and naturally) pick the professors whose opinion of him is best to write letter, so it's entirely possible that both students' letters say something like "The best student this year". Now of course this oversimplified analysis fails at a place like MIT, which graduated 85 physics majors last year, but the point is that a letter that seems quite strong at first look is merely average among admitted students.

The very best letters I have seen describe a student in some depth, including strengths and weaknesses. Including negatives actually helps the student (provided they are not too negative of course), because it shows that the writer isn't just writing fluff - she put time, effort and thought into the process, and it really can help the committee assess whether or not the student is a good match for the program. The more specific, the better. "Got an A in my class" but not much else isn't very helpful - we have the transcripts. "Good in labs but sometimes makes careless mathematical errors" is better. "Works well with ultrahigh vacuum equipment, and in fact has better vacuum hygiene than most postdocs, but still struggles with sign errors when doing lengthy matrix manipulation" is better still.

So, who should write your letters? The professors who know you the best. Those are not necessarily the biggest names at your university, or even necessarily the ones who gave you the highest grade. A detailed letter than is mostly, but not universally positive will do your application far more good than one that is completely positive but vague.

This is one of the areas where research is important. If you've done undergraduate research, you've worked closely with a professor, who can presumably write a letter with some meat on it. I would even argue that much of the benefit of undergraduate research on graduate admissions stems from the project generating a professor who can write such a letter. If you have not done any undergraduate research, I would strongly recommend having one letter from the professor teaching a laboratory course. Chances are she has interacted with you one-on-one, which is a plus and the admissions committee will also want to know how you did in the closest thing to research in your degree program.

If you have done something outside your own school, such as an REU, that is also a good source for letters: apart from the reasons above, now the committee knows what people at two schools think of you. It may make sense to have a professor in another department write you a letter, particularly if she knows you and your work well. Don't go overboard, though - if a physics major intending to get a PhD in physics sends in three letters from historians, the committee will wonder. Two physicists and a chemist though would not be a problem, and may be advantageous.

 

[Vanadium 50]

Part 5: Other Factors

Having experience with research at the undergraduate level is a good thing. There are people who claim that it is required to get into graduate school. I disagree. Beneficial, yes. Required, no.

One major benefit was mentioned earlier - it gives a professor an opportunity to work with you and write a letter with some substance to it. But what if you went to a small liberal arts college where research opportunities are limited? I wouldn't worry about it - most colleges that offer degrees in physics fall into that category, so you are hardly in an unusual situation. Many students are admitted with this sort of background, and they usually do quite well.

If however, you have an opportunity as an undergraduate to participate in research, you should certainly take it - there are personal benefits to this, and frankly, research isn't for everyone. If you find it's not for you, better to learn that as an undergraduate rather than after beginning a multi-year research degree. Also, it looks quite strange if one graduates from a research university, particularly one with a commitment to undergraduate involvement, with no research experience and then applies for a multi-year research program.

Often a candidate is asked to write a personal statement. This is not a contest to see who can write the saddest story or who was interested in physics the earliest. The committee doesn't care what books or television shows first got you interested in physics. They do, however, want to know why you want to invest half a dozen years of your life into this. They want to know what you want to study: experimental? theoretical? AMO? Nuclear? If your background is missing something typical of entering students (e.g. you were not a physics major as an undergrad), they want to know how you intend to make up that shortfall.

It's not expected that you have decided on your thesis topic at this point. But it is expected that you are aware of the different branches and have thought about where you might want to do your research. They are looking for something like "theoretical nuclear physics" and not "a better calculation of the half-life of Ni-56". If you are attracted by more than one area, say that. But if all branches of physics interest you equally, you might want to think a little harder.

Finally, for heaven's sake run this through a spell checker and look at the grammar. This is an opportunity to look very bad in front of the committee, and sadly, many students avail themselves of this opportunity.

 

[Defennder]

Part 2: Grades:
I don't know what the average GPA is of an admitted student, integrated over all universities, but I would imagine it's around 3.7: the typical student got mostly A's and some B's as an undergraduate. The less competitive one's undergraduate institution is, the higher the expectation of good grades. Below 3.5, a student starts to become uncompetitive very quickly. Below a 3.0 many universities simply will not admit you.

Just a question here, is this based on a 4.0 or 5.0 GPA scale?

And I believe this thread, once finalised would make a worthy addition to Zz's guide.

 

[Vanadium 50]

I was thinking mostly in terms of a 4.0 (which is the most common among undergraduate institutions).

 

[L62]

gradschoolshopper is a site that just links to that aip data. Any data that I have seen that claims a single digit rate is suspect. For example, USC claims they accept 13 out 190 but have 78 grad students. Rochester seems to claim they accept 20 out of 400 but have 114 grad students. They are either flat out lying (cooking the books or they honestly believe accepted students means students who accepted their offers) or have a 100% yield which neither Stanford, MIT , nor Harvard do. Dont take numbers at face value.

It could be that in the case of for example, USC - saying they accept 13 out of 190 but have 78 grad students - it's because the other 65 grad students were those who had been admitted in previous years who are still there working on their degrees. so the 13 out of 190 refers to new or incoming students whereas the 78 refers to total number of students (incoming as well as existing)

 

[j93]

It could be that in the case of for example, USC - saying they accept 13 out of 190 but have 78 grad students - it's because the other 65 grad students were those who had been admitted in previous years who are still there working on their degrees. so the 13 out of 190 refers to new or incoming students whereas the 78 refers to total number of students (incoming as well as existing)

I knew that was what those numbers referred too.
That would assume that USC has a 100% yield which I do not believe they do since no school has a 100% yield. I believe they mean students who accepted their offers which is equivalent to size of the new class but when I think of the word "accept" I think the amount of offers USC makes .I also believe that most people think the same including Harvard, Berkeley from their AIP data. I also believe that USC knows that too and are being dishonest to appear more selective.

 

[AstroChick]

I think it's a matter of being inaccurate rather than dishonest. I think the AIP sends out a form every year to the departments and the department secretaries have to fill it out. At least that was the case at my former school, which never took the form too seriously (but then again the department was totally backwards). I don't think anyone sits there and calculates the exact average of test scores and GPAs... Who has time for that?

 

[j93]

For Avg GPAs and GRE I would agree with you. I think if I was a secretary or anyone in the position to fill out the form and I received a form that asked about acceptances for my college I would assume they meant offers given by my university just like if they asked how many rejections I would think of the group that does not get an offer. I thinks it takes a deliberate effort to go against this interpretation especially since the AIP also asked for the amount of first year grad students.

 

[Vanadium 50]

I don't think that the exact number of rejected applications (which of course varies from school to school and year to year) is really that important. One very good reason is that there's not much an applicant can do about the other applications anyway, so it's best to focus on the one application they have some control over - their own. Another is that if the school accepts, say 20 students, it only matters if you're in that 20 or not. If not, it doesn't matter if you're in that batch with 5 other people or 500.

What matters is that even at a school ranked towards the bottom of PhD granting institutions (and these are often still quite good schools - the vast majority do not offer the PhD degree at all) there are many more applicants than places for them. Things are competitive everywhere, and like I said, not everyone who wants to go to graduate school gets to go.

 

[j93]

Just mentioned rejected applications because when you say rejected applications you mean applications that were not offered admission I am assuming and I believe that implies that when you say accepted you mean applications that were offered admission. USC and Rutgers apparently disagree with those definitions from the data they submitted to AIP and I can't believe they honestly do. The whole debate was to point out that physics PhD programs do not have single digit acceptance rate. The acceptance rate bottoms out at approximately 12% and can hover as high as 30% and slightly higher for domestic students. I was looking at UCLA data for domestics which is among top 50 programs. The rate for some lower ranked schools could possibly have acceptance rate in the high 30's/low 40's assuming they are at least slightly less selective than UCLA. That's a range from 1 in 8 to 1 in 3. This is according to AIP data that makes sense because it doesn't display a 100% yield and other university data. I just thought it was an exaggeration to imply a 5% acceptance rate.

 

[JUICYWART]

Just mentioned rejected applications because when you say rejected applications you mean applications that were not offered admission I am assuming and I believe that implies that when you say accepted you mean applications that were offered admission. USC and Rutgers apparently disagree with those definitions from the data they submitted to AIP and I can't believe they honestly do. The whole debate was to point out that physics PhD programs do not have single digit acceptance rate. The acceptance rate bottoms out at approximately 12% and can hover as high as 30% and slightly higher for domestic students. I was looking at UCLA data for domestics which is among top 50 programs. The rate for some lower ranked schools could possibly have acceptance rate in the high 30's/low 40's assuming they are at least slightly less selective than UCLA. That's a range from 1 in 8 to 1 in 3. This is according to AIP data that makes sense because it doesn't display a 100% yield and other university data. I just thought it was an exaggeration to imply a 5% acceptance rate.

While some top schools (I'm speaking as a Statistics PhD applicant) have slightly higher acceptance rates (such as Duke), generally, most students that apply to these schools are the best in the country [edit - best in the world] (think top 10%). So it doesn't really matter what the acceptance rate is. It's not a good indicator of how difficult it is to get into a graduate school. If you're an average applicant, your chance of getting into a top program will be MUCH less than 5%.

 

[j93]

While some top schools (I'm speaking as a Statistics PhD applicant) have slightly higher acceptance rates (such as Duke), generally, most students that apply to these schools are the best in the country [edit - best in the world] (think top 10%). So it doesn't really matter what the acceptance rate is. It's not a good indicator of how difficult it is to get into a graduate school. If you're an average applicant, your chance of getting into a top program will be MUCH less than 5%.

I am not a Statistics PhD applicant so I may not be as adept at statistical interpretation but
thats a pretty meaningless statement. Equivalently I could say that an adult with a 5th grade education/No community college courses has a nearly 0% chance of getting into his local state college. What relevance does this statement have to a college bound senior. Assuming someone is applying to a Physics PhD implies at least some degree of self selection, I doubt Harvard Physics gets applications from US History majors with a 2.3 UGPA and no PGRE. This self-selection is why acceptance rates have meaning to a physics grad especially given that its still admissions where nothing is completely deterministic.

 

[Troponin]

Thanks for taking the time to put this together Vanadium 50.

 

[uman]

I think this should be stickied, given the glut of "can I get in without a 3.0?" threads lately.

 

[kthouz]

Thank you Vanadium 50, this thread is very helpful for applicants.

 

[Voraldo]

How do you convert a percentage mark ie. 70% from a Canadian physics program into an American GPA? Is this 3.7 mark on a 4.0 or 4.33 scale?

On the other hand, where did you get your 3.7 gpa value from? It seems ridiculously high. :) The class averages of my physics and math classes at my university are usually around 72%.Thanks for your helpful post Vanadium50.

 

[Ryker]

If one's average was 70%, and the class average was 72%, I'd assume that person's GPA wouldn't be above 3.0, let alone 3.7.

 

[Fizex]

I know this is a year old but I have a question: Do grad schools tell their applicants if a TA or RA job is available to them after being accepted?

I'm also doubful on the below scenarios.

Situation 1: There was also a mention about some classes having more weight then others. What if an applicant had a 3.3 GPA but his college required him to take many humanities and social science courses which he did poorly in, but this student has aced every physics and math class he took. Would this make it very unlikely he would be accepted or does he have the grades that could make him a competitive applicant? Ceteris paribus.

Situation 2: How about an applicant with this upward trend of gpa's in his 4 years of undergrad: 2.6, 3.3, 3.7, 4.0. This gpa has an average of 3.4; would it be considered bad or good by a committee?

It seems that Vanadium has experience with acceptance committees so I would like people with similar experience to give an insight instead of speculation.

 

[jtbell]

I know this is a year old but I have a question: Do grad schools tell their applicants if a TA or RA job is available to them after being accepted?

Yes, when they offer admission they also tell you whether it includes a TA or RA, and the financial terms. At least that was the case when I applied to grad schools many years ago.

 

[Vanadium 50]

My experience is the same as JT Bell's.

As far as the other questions, the answer is, I am afraid, whatever the committee thinks of it. One school might look at low scores outside of physics and think "well, only his physics grades matter" and another might think "doesn't work so hard on things he's not interested in." That's why people get in in some places and don't in others.

 

[Choppy]

Volorado,
Most schools will have their own conversion schemes which should be printed in their calanders. For a very general approximation:
A+ = 4.0 = 90 - 100% (= 4.3)
A = 4.0 = 85 - 89%
A- = 3.7 = 80 - 84%
B+ = 3.3 = 77 - 79%
B = 3.0 = 73 - 76%
B- = 2.7 = 69 - 72%
etc.

In Canada, schools that have honour rolls will generally establish the cutoff around the 80%, A-, 3.7 line and the majority of students who get into graduate school are at or above this line.

Fizex,
Actually, most schools should be able to explain financial support before you even apply. It should be on their web pages. In some cases though, they won't make any guarantees until you receive a letter of offer.

For both of your scenarios, remember that graduate school admissions work on a competative basis. Once you make the minimum requirements, you are lumped into a pool of candidates for a set number of positions. Candidates in the pool are ranked and if there are N positions, the top N candidates are offered admission. So, in light of that, in scenario 1, this candidate would likely come out ahead of another candidate with the same average who didn't do as well in the upper year physics classes. Similarly, in scenario 2, this candidate would likely be ranked higher than one with the same average with consistent numbers or worse, a trend that went the other way.

 

[twofish-quant]

I think that the odds of getting into grad school if you are a serious student is a bit larger than those numbers indicate. The GRE is an international test so there are pretty substantial numbers of people taking it that will not end up in a US grad school.

There may be a lot of self-selection here, but every US citizen that I know that wanted to go to physics grad school with a decent application has gotten in somewhere, and I don't know anyone that has made a "serious application" that wasn't able to get in somewhere eventually.

 

[Jokerhelper]

I think that the odds of getting into grad school if you are a serious student is a bit larger than those numbers indicate. The GRE is an international test so there are pretty substantial numbers of people taking it that will not end up in a US grad school.

There may be a lot of self-selection here, but every US citizen that I know that wanted to go to physics grad school with a decent application has gotten in somewhere, and I don't know anyone that has made a "serious application" that wasn't able to get in somewhere eventually.

Is it? I thought only US grad schools wanted those.

 

[Caramon]

Volorado,
Most schools will have their own conversion schemes which should be printed in their calanders. For a very general approximation:
A+ = 4.0 = 90 - 100% (= 4.3)
A = 4.0 = 85 - 89%
A- = 3.7 = 80 - 84%
B+ = 3.3 = 77 - 79%
B = 3.0 = 73 - 76%
B- = 2.7 = 69 - 72%

In Alberta from my experience it generally goes like this:
A+ = 4.0 = 97% +
A = 3.9 = 93%-96%
A- = 3.7 = 90%-92%
B+ = 3.3 = 85%-89%
B = 3.0 = 80% - 84%
B- = 2.7 = 75%-79%
C+ = 2.3 = 70%-74%
C = 2.0 = Below 70%

There is no "set" percentage, it's based on z-scores and a bell-curve normally.
Not sure how the hell someone would be worth any of A with a grade in the "80-84" range...

 

[Choppy]

Hi Camaron,

Here's a conversion chart from McMaster's website. As you can see, it's pretty school-dependent. Also, there's a difference between percentage obtained on exams and final grades. The 3.7 = A- = 80-84% line seems pretty standard from my experience. It's also worth pointing out that this is for undergrad. My experience is that graduate grades, although following a similar scale, will have a significantly higher cutoff for what constitutes a pass.

http://careers.mcmaster.ca/students/education-planning/virtual-resources/gpa-conversion-chart

 

[Ryker]

In Alberta from my experience it generally goes like this:
A+ = 4.0 = 97% +
A = 3.9 = 93%-96%
A- = 3.7 = 90%-92%
B+ = 3.3 = 85%-89%
B = 3.0 = 80% - 84%
B- = 2.7 = 75%-79%
C+ = 2.3 = 70%-74%
C = 2.0 = Below 70%

There is no "set" percentage, it's based on z-scores and a bell-curve normally.
Not sure how the hell someone would be worth any of A with a grade in the "80-84" range...

Where did you get 3.9 from? Judging by the UofA's official grading system explanation, an A translates to a 4.0. Also, I know I just finished my first semester here, but from my experience thus far mark boundaries are lower. For example, in Newtonian Mechanics you were able to get an A+ with less than 90% total, and the cut-off point for an A- in some of the Maths courses was as low as 83%, as well. I can't really tell for other courses, though, because I didn't see their mark distribution.

 

[twofish-quant]

Is it? I thought only US grad schools wanted those.

I know that some schools in China encourage people to take the GRE. In any case, if you are even thinking about applying to a US school, you'll need to take the GRE, even if you don't end up going. Also US universities generally give some preference to US citizens, and native English speakers don't have to worry about TOEFL scores.

So if you are a US citizen, I think your chances of getting in somewhere are considerably higher than 50/50.

Also one important thing about GPA is that there is a threshold effect. If you have a low <3.0 GPA, then you will find it extremely, extremely difficult to get in. On the other hand, the difference between a 3.7 and 3.8 is pretty much irrelevant. The reason for this is that schools are different enough so that it's pretty much impossible to compare a 3.7 and 3.8, but if you have a 2.9, then you really did mess up somewhere along the line.