# A survey of equal sums of like powers

> An annotated edition of this paper on Fermat's Library, where scientists and readers explain its ideas, equations and history in the margins. Annotated edition: https://fermatslibrary.com/s/a-survey-of-equal-sums-of-like-powers

- **Authors:** Lander, Parkin and Selfridge
- **Original:** https://s3-us-west-1.amazonaws.com/fermatslibrary/uploads%2F1465885143681-uml7gp62auvybsn1-69194c5a15c6b6d7f64f2baad1f16fbc%2Flander1967.pdf
- **Pages:** 14
- **Annotations:** 0 comments in 0 threads
- **Featured:** Fermat's Library Journal Club
- **Annotated edition:** https://fermatslibrary.com/s/a-survey-of-equal-sums-of-like-powers

## Full text

Extracted from the PDF. [Annotation N] marks where each annotation above sits.

#### Page 1

A Survey of Equal Sums of Like Powers

By L. J. Lander, T. R. Parkin and J. L. Selfridge

Introduction. The Diophantine equation

(1) Xl + X2k + + Xmk =Yi + Y2 + + Ynk < m < n, has been studied by numerous mathematicians for many years and by various methods [1], [2]. We recently conducted a series of computer searches using the CDC 6600 to identify the sets of parameters k, m, n for which solutions exist and to find the least solutions for certain sets. This paper outlines the results of the computation, notes some previously published results, and concludes with a table showing, for various values of k and m, the least n for which a solution to (1) is known. We restrict our attention to k < 10. We assume that the xi and yj are positive integers and xi # yj. We do not distinguish between solutions which differ only in that the xi or yj are rearranged. We will refer to (1) as (k. m. n) and say that a primitive solution to (k. m. n) is one in which no integer > 1 divides all the numbers X1, X2, *, Xm, Yi, Y2, , yn. Putting

m n Z= E i = E 1 1 we order the primitive solutions according to the magnitude of z and denote the rth primitive solution to (k. m. n) by (k. m. n) r. Where we refer to the range covered in a search for solutions, we mean the upper limit on z. The notation (Xl, X2, ., xm)k = (yl, y2, , yn)k means X - = yjk. Any parametric solution discussed does not include all solutions unless otherwise stated.

Squares and Cubes. For k = 2 the general solution of the Pythagorean equation (2. 1. 2) is well known [3]. Many solutions in small integers and various parametric solutions have been given for (2. 1. n) wi'th n > 3. The general solution of (2. 2. 2) is known [4]. Solutions to (2. 2. n) with n > 3 and (2. m. n) with m > 3 are numerous. The impossibility of solving (k. 1. 2) with k > 3 is Fermat's last theorem, which has been established for k < 25000 [5]. The general solution of (3. 1. 3) in rationals is attributed to Euler and Vieta [6] and also produces all solutions to (3. 2. 2) if the arguments are properly chosen. There are many solutions in small integers and various parametric solutions to (3. 1. n) with n > 4 and to (3. m. n) with m > 2 [7].

Fourth Powers. (4. 1. n)-For n = 3, no solution is known. M. Ward [8] developed congruential constraints which, together with some hand computing, allowed him to show that x4 = Yi4 + Y24 + Y34 has no solution if x < 10,000. The authors extended the search on the computer using a similar method and verified that there is no solution for x < 220,000. Ward showed that if x4 = Y14 + Y24 + y34is a primitive solution, it may be assumed that x, y =_ 1 (mod 2), Y2, Y3 =0 (mod 8) and either x - yi or x + yi

Received September 30, 1966. Revised December 12, 1966. 446

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 2

A SURVEY OF EQUAL SUMS OF LIKE POWERS 447

is -0 (mod 1024). Also x 9 0 (mod 5) or else all yi would be 0 (mod 5) since U4 _ 0 or 1 according as u -0 or u 0 0 (mod 5). The computer program generated all numbers M = (x4 - y4)/2048 with 0 < yi < x, x prime to 10 and Yi =- x (mod 1024). Tests were applied to M = (y2/8)4 + (y3/8)4 to reject cases in which a solution would not be primitive or M could not be the sum of two biquadrates. If M passed all the tests, its decomposition was attempted by trial using addition of entries in a stored table of biquadrates (27500 entries for x < 220,000 = 8 .27500). The tests were:

(1) M must be = 0, 1 or 2 (mod 16) and (mod 5); (2) M must not be =7, 8 or 11 (mod 13) and must not be =4, 5, 6, 9, 13, 22 or 28 (mod 29); (3) x and yi must not both be divisible by an odd prime p = 3, 5 or 7 (mod 8) for if so, p4 divides M, p divides Y2 and y3 and the solution is not primitive; (4) M must not have a factor p where p is an odd prime not = 1 (mod 8) unless p4 also divides M. In this case p divides y2 and y3, and in the decomposition by trial M can be replaced by M/p4 (here tests were made only for p < 100). Of approximately 19,200,000 initial values of M, only 22,400 required the trial decomposition.

TABLE I Primitive solutions of (4. 1. 4) for z < (8002)4 zx4= >4 yj4 Z Xi Zy

j Xi Yi Y2 Y3 Y4 Ref.

1 353 30 120 272 315 [9] 2 651 240 340 430 599 [34] 3 2487 435 710 1384 2420 [10] 4 2501 1130 1190 1432 2365 [10] 5 2829 850 1010 1546 2745 [10]

6 3723 2270 2345 2460 3152 [10] 7 3973 350 1652 3230 3395 [10] 8 4267 205 1060 2650 4094 [10] 9 4333 1394 1750 3545 3670 10 4449 699 700 2840 4250

11 4949 380 1660 1880 4907 12 5281 1000 1120 3233 5080 13 5463 410 1412 3910 5055 14 5491 955 1770 2634 5400 [11] 15 5543 30 1680 3043 5400

16 5729 1354 1810 4355 5150 17 6167 542 2770 4280 5695 18 6609 50 885 5000 5984 19 6801 1490 3468 4790 6185 20 7101 1390 2850 5365 6368

21 7209 160 1345 2790 7166 22 7339 800 3052 5440 6635 23 7703 2230 3196 5620 6995

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 3

448 L. J. LANDER, T. R. PARKIN AND J. L. SELFRIDGE For n = 4, R. Norrie [9] found the smallest solution (353)4 = (30, 120, 272, 315)4. J. 0. Patterson [34] found (4. 1. 4)2 and J. Leech [10] found the next 6 primitive solutions on the EDSAC 2 computer. S. Brudno [11] gave another primitive solution, the 14th in our Table I. The authors exhaustively searched the range 80024 using Leech's method finding in all the 23 primitives listed in Table I. No parametric solution has been found for (4. 1. 4) although the general solution is known for (3. 1. 3) and a parametric solution (discussed later) is known for (5. 1. 5). TABLE II Primitive solutions of (4. 2. 2) for 7. 5 X 1015 < z < 5. 3 X 1016 Z = X14 + X24 = Y14 + Y24 i Xl X2 Yi Y2 Z *32 6262 8961 7234 8511 7 98564 45223 00177 33 5452 9733 7528 9029 9 85755 13638 85937 34 3401 10142 7054 9527 10 71400 42234 80497 35 5277 10409 8103 9517 12 51457 36160 92402 36 3779 10652 8332 9533 13 07827 22453 98097 37 3644 11515 5960 11333 17 75781 85225 58321 38 1525 12234 3550 12213 22 40674 37332 52161 **39 2903 12231 10203 10381 22 45039 16406 17602 40 1149 12653 7809 12167 25 63324 34950 11682 41 5121 13472 9153 12772 33 62808 84147 85537 42 5526 13751 11022 12169 36 68751 70593 08977 43 6470 14421 8171 14190 45 00187 64129 98081 44 6496 14643 11379 13268 47 75551 49900 03857 45 261 14861 8427 14461 48 77442 72266 31682 46 581 15109 8461 14723 52 11273 11403 26882 * For solutions to (4.2.2) for i = 1 to 31 see Lander and Parkin [18]. ** This solution was found by Euler [371. For n > 5 there exist many solutions in small integers. (4. 1. 5)1 is (5)4 = (2, 2, 3, 4, 4)4. Several parametric solutions to (4. 1. 5) are known due to E. Fauquembergue [12], C. Haldeman [13], and A. Martin [14]. (4. 2. n)-For n = 2 the least solution is (59, 158)4 = (133, 134)4. Euler [15] gave a two-parameter solution and A. Gerardin [16] gave an equivalent but simpler form of this solution. Several of the smaller primitive solutions were found by Euler, A.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC All use subject to JSTOR Terms and Conditions

#### Page 4

A SURVEY OF EQUAL SUMS OF LIKE POWERS 449

Werebrusow, and Leech [17] and a recent computer search by Lander and Parkin [18] extended the list of known primitives to 31. More recently we have increased this to a total of 46 primitives by a complete search of the range 5.3 X 1016 and the 15 new primitives are listed in Table II. The general solution is not known. For n > 3 there are many small solutions. (4. 2. 3)1 is (7, 7)4 = (3, 5, 8)4. Several parametric solutions are known for (4. 2. 3) due to Gerardin [19] and F. Ferrari [20]. (4. m. n)-For m > 3, solutions in small integers are numerous. Parametric solutions to (4. 3. 3) were given by Gerardin [21] and Werebrusow [22]. (4. 3. 3)1 is (2, 4, 7)4= (3, 6, 6)4.

Fifth Powers. (5. 1. n)-For n = 3, no solution is known. Lander and Parkin [23], [24] found (5. 1. 4), to be (144)5 = (27, 84, 110, 133)5. This disproved Euler's conjecture [25] that (k. 1. n) has no solution if 1 < n < k. No further primitive solutions to (5. 1. 4) exist in the range up to 7655. Form = 5, S. Sastry and S. Chowla [26] obtained a two-parameter solution yielding (107) 5 (7, 43, 57, 80, 100)5 as its minimal primitive; this solution is (5. 1. 5)>. Lander and Parkin [24] found (5. 1. 5) and (5. 1. 5)2 to be (72)5 = (19,43,46,47, 67)5 and (94)5 = (21, 23, 37, 79, 84)5. More recently we searched the range up to 5995 and found in all the twelve primitive solutions given in Table III.

TABLE III Primitive solutions of (5. 1. 5) for z < 5995 z =x15 = E5 yj5

i Xj Yi /2 13 14 Y5 Ref.

1 72 19 43 46 47 67 [24]

2 94 21 23 37 79 84 [24]

3 107 7 43 57 80 100 [26]

4 365 78 120 191 259 347

5 415 79 202 258 261 395

6 427 4 26 139 296 412

7 435 31 105 139 314 416

8 480 54 91 101 404 430

9 503 19 201 347 388 448

10 530 159 172 200 356 513

11 553 218 276 385 409 495

12 575 2 298 351 474 500

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 5

450 L. J. LANDER, T. R. PARKIN AND J. L. SELFRIDGE

For n > 6 there are solutions in moderately small integers. (5. 1. 6)1 is (12)5 = (4, 5, 6, 7, 9, 11)5 found by A. Martin [27]. The first eight primitive solutions to (5. 1. 6) are given in [24]. (5. 1. 7), is (23)5 = (1, 7, 8, 14, 15, 18, 20)5. (5. 2. n)-No solution is known for n < 3. An exhaustive search by the authors verified that there is no solution to (5. 2. 2) in the range up to 2.8 X 1014 or to (5. 2. 3) in the range up to 8 X 1012. Sastry's parametric solution for (5. 1. 5) mentioned above gives for certain values of its arguments solutions to (5. 2. 4), the smallest being (12, 38)5 = (5, 13, 25, 37)5 which is (5. 2. 4)2. K. Subba Rao [28] found (3, 29)5 = (4, 10, 20, 28)5 which is (5. 2. 4)1. Table IV lists the ten primitives which exist in the range up to 2 X 1010.

TABLE IV Primitive solutions of (5. 2. 4) for z < 2 X 1010 Z = Ylxjl - ' yj5

i Xi X2 Yi Y2 Y3 Y4 z Ref.

1 3 29 4 10 20 28 205 11392 [28]

2 12 38 5 13 25 37 794 84000 [26]

3 28 52 26 29 35 50 3974 14400

4 61 64 5 25 62 63 19183 38125

5 16 85 6 50 53 82 44381 01701

6 31 96 56 63 72 86 81823 56127

7 14 99 44 58 67 94 95104 38323

8 63 97 11 13 37 99 95797 76800

9 25 106 48 57 76 100 1 33920 21401

10 54 111 58 76 79 102 1 73097 46575

For n > 5 there are solutions in moderately small integers; (5. 2. 5), is (1, 22)5= (4, 5, 7, 16, 21)' due to Subba Rao [28]. We give the first six primitives for (5. 2. 5) in Table V.

(5. 3. n)-The first solution known for n = 3 was (49, 75, 107)5 = (39, 92, 100)5 due to A. Moessner [35]; this is (5. 3. 3)5. H. P. F. Swinnerton-Dyer gave two separate two-parameter solutions [36]. We give the 45 primitives in the range up to 8 X 1012 in Table VI. For n > 4, solutions in small integers are plentiful. (5. 3. 4), is (3, 22, 25)5 = (1, 8, 14, 27)5 due to Subba Rao [28]. A two-parameter solution to (5. 3. 4) was given by G. Xeroudakes and A. Moessner [29]. (5. m. n)-If m > 4, there are many solutions in small integers. (5. 4. 4)1 is (5, 6, 6, 8)5 = (4, 7, 7, 7)5 due to Subba Rao [28]. Several parametric solutions to (5. 4. 4) were found by Xeroudakes and Moessner [29]. The first triple coincidence of four fifth powers is 1479604544 = (3,48,52, 61)5 = (13,36,51, 64)5 = (18,36,44,66)5.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 6

A SURVEY OF EQUAL SUMS OF LIKE POWERS 451

In the subsequent discussion we adopt a notation borrowed from the field of partitions, writing xr to signify the term x repeated r times in the expression in which it appears. Table VII uses this notation, giving (k. m. n) 1 where known and references solutions in other tables. Table VII also shows for certain (k. m. n) the range which has been searched on the computer exhaustively. For the remainder of the equations (k. m. n) which are discussed we note in the text only the limits searched, interesting features, and methods employed; specific solutions are given in Table VII.

Sixth Powers. (6. 1. n)-No solution is known for n < 6. We consider the cases of n = 6, 7 and 8 in descending order. To solve (6. 1. 8), x6 I > yi6, note that u6 0 or 1 (mod 9) according as u 0 or u 5 0 (mod 3). Then if x 0 (mod 3), all y 0 (mod 3) and the solution is not primitive. Therefore take x and exactly one of the yi (say yi) prime to 3. Then (x6 - yi6)/36 = 28 (yi/3)6 is an integer (which is true if and only if yi _ ix (mod 243)) to be decomposed by trial as the sum of 7 sixth powers. In Table VIII we give the 14 smallest primitives found by this method; (6. 1. 8)1 is (251)6 = (8, 12, 30, 78, 102, 138, 165, 246)6.

TABLE V Primitive solutions of (5. 2. 5) for z < 2. 8 X 108 Z = 2 xjl= - yj

i X1 X2 Yi Y2 Y3 Y4 l5 Z

*1 1 22 4 5 7 16 21 51 53633

2 23 29 9 11 14 18 30 269 47492

3 16 38 10 14 26 31 33 802 83744

4 24 42 4 22 29 35 36 1386 53856

5 30 44 8 15 17 19 45 1892 16224

6 36 42 5 6 26 27 44 1911 57408

* The first solution is due to Subba Rao [28].

For (6.1. 7), X6 = 11 yi6, note that u6 = 0 or 1 (mod 8) according as u is even or odd. Then for a primitive solution, x and exactly one of the yi are odd. The argument for (6. 1. 8) modulo 9 applies and x is prime to 6, y, (say) is prime to 3, and either y, is odd or another y (say y2), is odd. In the first case y =_ ix (mod 243) and (mod 32) and (X6 - y16)/66 = Z2 (yi/6)6 is an integer to be decomposed by trial as the sum of 6 sixth powers. In the second case yi ?x (mod 243), Y2 34?X (mod 32) and (X6 - y16 - y26)/66 = 7 (yi/6)6 must be an integer (certain combinations X, Yl, Y2 satisfying the congruences are rejected) which is decomposed by trial as the sum of 5 sixth powers. The only solution for x < 1536 is (6. 1. 7)1, (1141)6 = (74, 234, 402, 474, 702, 894, 1077)6 which is obtained in the second case.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 7

452 L. J. LANDER, T. R. PARKIN AND J. L. SELFRIDGE

TABLE VI

Primitive solutions of (5. 3. 3) for z < 8 X 1012

z Exj5 = -3 yj5

i Yi[ XI X2 YX3 i Y2 Y3 Z

1 24 28 67 3 54 62 13752 98099 2 18 44 66 13 51 64 14191 38368 3 21 43 74 8 62 68 23700 99168 4 56 67 83 53 72 81 58398 97526 *5 49 75 107 39 92 100 1 66810 39431 6 26 85 118 53 90 116 2 73265 12069 7 38 47 123 1 89 118 2 84616 37018 8 73 96 119 68 106 114 3 40903 35168 9 39 56 136 3 97 131 4 71668 30151 10 13 35 142 17 95 138 5 77882 32400 11 28 32 155 91 94 150 8 95168 61675 12 65 94 152 42 129 140 8 96361 42881 13 63 67 169 9 131 159 14 02010 53499 14 68 137 170 36 140 169 19 17013 58025 15 43 109 181 13 159 161 20 97974 92893 16 74 113 182 61 129 179 22 03336 44849 17 39 142 186 28 167 172 28 04458 41607 18 44 55 201 18 152 190 32 87486 01600 19 58 101 204 113 145 195 36 44723 14293 20 18 31 215 10 183 191 45 94319 03094 21 19 168 216 11 183 209 60 40152 82243 22 5 145 224 153 157 214 62 80466 82374 23 27 106 229 12 122 228 64 31599 96832 24 151 166 233 126 208 216 89 12718 82720 25 59 139 248 23 184 239 99 07237 88966 26 157 193 234 147 218 219 106 47575 48174 27 2 97 258 35 125 257 115 17249 93057 28 3 121 264 163 185 250 130 83259 82668 29 97 181 274 67 227 258 174 72267 67782 30 99 105 286 30 179 281 193 57802 02300 31 132 154 283 80 219 270 194 19238 97099 32 106 137 288 201 219 261 204 29996 35401 33 40 168 289 3 215 279 214 99241 22017 34 136 158 294 71 249 268 234 15192 15168 35 193 229 282 179 259 266 268 09353 50774 36 107 229 293 93 259 277 280 32137 94149 37 31 173 307 7 201 303 288 20348 39551 38 102 118 310 49 270 271 289 68334 85600 39 116 124 310 21 235 294 291 32347 67200 40 30 39 331 65 224 321 397 33103 34850 41 119 232 328 89 289 301 449 23488 61399 42 108 181 348 53 246 338 531 27877 53637 43 114 211 364 52 298 339 682 75705 13699 44 172 206 364 102 303 337 691 15935 15232 45 123 137 373 13 259 361 729 65305 14393

* This solution was found by A. Moessner [35].

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 8

A SURVEY OF EQUAL SUMS OF LIKE POWERS 453

TABLE VII (k. m. n)i and summary of results

Range (k. m. n) Searched Solutions Known*

4. 1. 3 2.34 X 1021 None known 4. 1. 4 4.1 X 1015 (353)4 = (30, 120, 272, 315)4 See Table I, 23 solutions 4. 1. 5 (5)4 = (223, 42)4 4. 2. 2 5.3 X 1011 (59, 158)4 = (133, 134)4 See Table I in [18], and Table II, 46 solutions 4. 2. 3 (72)4 = (3, 5, 8)4 4. 3. 3 (2, 4, 7)4 = (3, 62)4 5. 1. 3 2.6 X 1014 None known 5. 1. 4 2.6 X 1014 (144)5 = (27, 84, 110, 133)5 5. 1. 5 7.7 X 1013 (72)5 = (19, 43, 46, 47, 67)5 See Table III, 12 solutions 5. 1. 6 (12)5 = (4, 5, 6, 7, 9, 11)5 5. 1. 7 (23)5 = (1, 7, 8, 14, 15, 18, 20)5 5. 2. 2 2.8 X 1014 None known 5. 2. 3 8 X 1012 None known 5. 2. 4 2 X 1010 (3, 29)5 = (4, 10, 20, 28)5 See Table IV, 10 solutions 5. 2. 5 2 X108 (1, 22)= (4, 5,7,16, 21)5 See Table V, 6 solutions 5. 3. 3 8 X 1012 (24, 28, 67)5 = (3, 54, 62)5 See Table VI, 45 solutions 5. 3. 4 (3, 22, 25)5 = (1, 8, 14, 27)5 5. 4. 4 (5, 62, 8)5 = (4, 73)5 6. 1. n 3.16 X 1027 None known for n < 6 6. 1. 7 1.3 X 1019 (1141)6 = (74, 234, 402, 474, 702, 894, 1077)6 6. 1. 8 5.8 X 1016 (251)6 = (8, 12, 30, 78, 102, 138, 165, 246)6 See Table VIII, 14 solutions 6. 1. 9 (54)6 = (1, 17, 19, 22, 31, 372, 41, 49)6 6. 1. 10 (39)6 = (2, 4, 7, 14, 16, 262, 30, 322)6 6. 1. 11 (18)6 = (2, 53, 72, 92, 10, 14, 17)6 6. 2. n 4 X 1012 None known for n < 6 6. 2. 7 (56, 91)6 = (18, 22, 36, 58, 69, 782)6 6. 2. 8 (35, 37)6 = (8, 10, 12, 15, 24, 30, 33, 36)6 6. 2. 9 (6, 21)6 = (1, 52, 7, 133, 17, 19)6 6. 2. 10 (122)6 = (13, 42, 7, 9, 1 13) 6. 3. 3 2.5 X 1014 (3, 19, 22)6 = (10, 15, 23)6 See Table IX, 10 solutions 6. 3. 4 2.9 X 1012 (41, 58, 73)6 = (15, 32, 65, 70)6 See Table X, 5 solutions 6. 4. 4 (22, 92)6 = (3, 5, 6, 10)6 7. 1. n 1.95 X 1014 None known for n < 7 7. 1. 8 (102)7 = (12, 35, 53, 58, 64, 83, 85, 90)7 7. 1. 9 (62)7 = (6, 14, 20, 22, 27, 33, 41, 50, 59)7 7. 2. 8 (10, 33)7 = (5, 6, 7, 152, 20, 28, 31)7 7. 3. 7 (26, 302)7 = (72, 12, 16, 27, 28, 31)7 7. 4. 5 (12, 16, 43, 50)7 = (3, 11, 26, 29, 52)7 7. 5. 5 (82, 13, 16, 19)7 = (2, 12, 15, 17, 18)7 See Table XI, 17 solutions

* All solutions shown are (k. m. n)i unless otherwise marked.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 9

454 L. J. LANDER, T. R. PARKIN AND J. L. SELFRIDGE

TABLE VII (cont.)

Range (k. m. n) Searched Solutions Known

7. 6. 6 (2, 3, 62, 10, 13)7 = (12, 72, 122)7 8. 1. 11 (125)8 (14, 18, 442, 66, 70, 92, 93, 96, 106, 112)8 8. 1. 12 (65)8 (82, 10, 243, 26, 30, 34, 44, 52, 63)8 8. 2. 9 (11, 27)8 = (2, 7, 8, 16, 17, 202, 242)8 8. 3. 8 (8, 17, 50)8 = (6, 12, 162, 382, 40, 47)8 8. 4. 7 (6, 11, 20, 35)8 = (7, 9, 16, 222, 28, 34)8 8. 5. 5 (1, 10, 11, 20, 43)8 = (5, 28, 32, 35, 41)8 8. 6. 6 (3, 6, 8, 10, 15, 23)8 = (5, 92, 12, 20, 22)8 8. 7. 7 (1, 3, 5, 62, 8, 13)8 (4, 7, 92, 10, 11, 12)8 8. 8. 8 (1, 3, 73, 102, 12)8 (4, 52, 62, 113)8 9. 1. 15 (26)9 = (22, 4, 62, 7, 92, 10, 15, 18, 212, 232)9 9. 2. 12 (15, 21)9 = (24, 32, 4, 7, 16, 17, 192)9 9. 3. 11 (13, 16, 30)9 = (2, 3, 6, 7, 92, 192, 21, 25, 29)9 9. 4. 10 (5, 12, 16, 21)9 = (2, 62, 9, 10, 11, 14, 18, 192)9 9. 5. 11 (7, 8, 14, 20, 22)9 = (3, 52, 92, 12, 152, 16, 212)9 9. 6. 6 (1, 132, 14, 18, 23)9 = (5, 9, 10, 15, 21, 22)9 10. 1. 23 (15)10 (11, 2, 3, 6, 76, 94, 10, 122, 13, 14)10 10. 2. 19 (9, 17)10 (25, 5, 6, 10, 116, 122, 153)10 10. 3. 24 (11, 152)10 = (1, 2, 3, 410, 7, 8 7, 10, 12, 16)10 10. 4. 23 (113, 16)10 (15, 22, 32, 4, 64, 73, 8, 102, 142, 15)10 10. 5. 16 (32 8 14, 16)10 = (14, 2, 42, 6 122, 135, 15)10 10. 6. 27 (22, 8 11,122)10 = (1, 34, 42, 52, 67, 79, 10, 13)10 *10. 7. 7 (1, 28, 31, 32, 55, 61, 68)10 = (17, 20, 23, 44, 49, 64, 67)10

* Moessner [35]; not known to be (10. 7. 7)1.

For (6. 1. 6), x6 = Z1 yt6 note that u6 0 or 1 (mod 7) according as u 0 or u q 0 (mod 7). Then for a primitive solution, x and exactly one of the y, (say yi) are prime to 7. This implies y, _ ix, iqx or ?q2x where q = 34968 is a primitive sixth root of unity (mod 76 = 117649). Now the foregoing arguments modulo 8 and modulo 9 apply, and there are five cases. (1) If yli ?41 (mod 6) then yl- -x (mod 243) and (mod 32) and (X6 - yi6)/426 = 6 (yj/42)6 is an integer to be decomposed by trial as the sum of 5 sixth powers.

(2) If y- -42 (mod 6) then y, _4ix (mod 243) and another of the yi (say Y2), is odd. Then Y2 0 (mod 3.7), Y2 a+x (mod 32), and (X6 - -y26)/426 = 36 (yi/42)6 is the sum of 4 integral sixth powers. (3) If y, 3 (mod 6) then y -4-x (mod 32) and another of the yi (say y2), is prime to 3, Y2=O (mod 2.7), and y2-4x (mod 243). In case (2), (X6 - y 16-y26)/426 is an integer and is the sum of 4 sixth powers. (4) If y-= 0 (mod 6), another of the yi (say y2), is prime to 3, Y2 0 (mod 7) and Y2 ?x (mod 243). If Y2 is odd, then Y2 ax (mod 32) and as in cases (2) and (3) (X6 - y6 - Y26)/426 is the sum of 4 sixth powers. If Y2 is even, we have case (5).

(5) Another of the yi (say Y3), is odd, y3 _ 0 (mod 3.7), y3 - 4x (mod 32), and (X6 - 16 - Y26 - y36)/426 = 6 (yi/42)6 is an integer to be decomposed as the sum of 3 sixth powers.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 10

A SURVEY OF EQUAL SUMS OF LIKE POWERS 455 The search for a solution to (6. 1. 6) was carried exhaustively by this method through the range x < 38314 and there is no solution in this range. A. Alartin [30] gave a solution to (6. 1. 16); Moessner [31] gave solutions to (6. 1. n) for n = 16, 18, 20 and 23. For n > 11, it is not difficult to find solutions in small inteaers.

TABLE VIII Primitive Solutions of (6. 1. 8) for z < 7 X 101O Z = = 8 yi6 i 'X Yi Y2 Y3 Y4 Y5 Y6 j7 Y8 1 251 8 12 30 78 102 138 165 246 2 431 48 111 156 186 188 228 240 426 3 440 93 93 195 197 303 303 303 411 4 440 219 255 261 267 289 351 351 351 5 455 12 66 138 174 212 288 306 441 6 493 12 48 222 236 333 384 390 426 7 499 66 78 144 228 256 288 435 444 8 502 16 24 60 156 204 276 330 492 9 547 61 96 156 228 276 318 354 534 10 559 170 177 276 312 312 408 450 498 11 581 60 102 126 261 270 338 354 570 12 583 57 146 150 360 390 402 444 528 13 607 33 72 122 192 204 390 534 534 14 623 12 90 114 114 273 306 492 592

(6. 3. n)-Subba Rao [32] found the solution (3, 19, 22)6 = (10, 15, 23)6 which is (6. 3. 3)1. In Table IX we give the remaining 9 primitive solutions which exist in the range up to 2. 5 X 1014. It is interesting to note that each of the solutions except the sixth is also a solution to (2. 3. 3). Table X gives the five primitive solutions to (6. 3. 4) which exist in the range up to 2. 9 X 1012. (6. m. n)-If m is > 4, solutions in small integers can be found readily. Subba Rao [32] gave (6. 4. 4) (see Table VII). The first triple coincidence of 4 sixth powers is 1885800643779 = (1, 34, 49, 111)6 = (7, 43, 69, 110)6 = (18, 25, 77, 109)6. Seventh Powers. (7. 2. 10)2 is (2, 27)7= (4, 8,13,142,16,18, 22, 232)7= (72, 9,13,14,18,20,222, 23)7 which is a double primitive and reduces to the solution (7. 5. 5)2.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC All use subject to JSTOR Terms and Conditions

#### Page 11

456 L. J. LANDER, T. R. PARKIN AND J. L. SELFRIDGE

TABLE IX Primitive solutions of (6. 3. 3) for z < 2.5 X 1014 z = 1j x6 = Z1 yj6

i X1 X2 X3 Yi Y2 Y3 Z

*1 3 19 22 10 15 23 1604 26514

2 36 37 67 15 52 65 9 52008 90914

3 33 47 74 23 54 73 17 62771 73474

4 32 43 81 3 55 80 28 98246 41354

5 37 50 81 11 65 78 30 06202 62890

6 25 62 138 82 92 135 696 38068 13393

7 51 113 136 40 125 129 842 70669 28346

8 71 92 147 1 132 133 1082 47536 54794

9 111 121 230 26 169 225 15304 47319 28882

10 75 142 245 14 163 243 22464 65092 02194

* The first solution is due to K. Subba Rao [32]. TABLE X Primitive solutions of (6. 3. 4) for z < 2.9 X 1012 Z = 3 Xjl = 41 yjl

i YX1 X2 X3 Yi 2 Y3 Y4 Z

1 41 58 73 15 32 65 70 19 41530 23074 2 61 62 85 52 56 69 83 48 54701 25570 3 61 74 85 26 56 71 87 59 28763 80162 4 11 88 90 21 74 78 92 99 58468 58345 5 26 83 95 23 24 28 101 106 23411 79770

(7. 5. n)-Table XI lists the 17 primitive solutions to (7. 5. 5) which exist in the range up to 4. 0 X 1012.

Eighth Powers. (8. 1. n)-We found a parametric solution to (8. 1. 17), (28k+4 + 1)8 = (28k+4 - 1)8 + (27k+4)8 + (2k+l)8 + 7[(25k+3)8 + (23k+2)8] which for k = 0 yields (8. 1. 17)1. This was the solution used by Sastry [26] in developing a parametric solution to (8. 8. 8). The computer program used in searching for solutions to (8. 1. n) was based on the congruences x8 =0 or 1 (mod 32) according as x _ 0 or 1 (mod 2) so that primitive solutionsto x8 = l y j8with n < 32 must have x and (say) y both odd. Then x8 - yi8 is divisible by 28 which implies x = --y (mod 32), and (X8 - y18)/256 is decomposed as the sum of n - 1 eighth powers by trial. Solutions to (8. 5. 5) and (8. 9. 9) were found by A. Letac [33].

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 12

A SURVEY OF EQUAL SUMS OF LIKE POWERS 457

Ninth and Tenth Powers. Computations performed by the authors for (9.m.n) and (10. m. n) are the basis for the data shown in the last two columns of Table XII,

TABLE XI Primitive solutions of (7. 5. 5) for z < 4.0 X 1012

z = x Xj =- y

i Xi X2 X3 X4 X,5 Yi Y2 Y3 Y4 Y5 z

1 8 8 13 16 19 2 12 15 17 18 12292 50016 2 4 8 14 16 23 7 7 9 20 22 37807 87943 3 11 12 18 21 26 9 10 22 23 24 1 05004 37728 4 6 12 20 22 27 10 13 13 25 26 1 42708 22835 5 3 13 17 24 38 14 26 32 32 33 11 94751 43393 6 4 5 30 36 44 2 8 27 39 43 41 95120 68269 7 16 33 33 33 44 18 26 34 38 43 44 74015 74051 8 3 4 21 39 45 14 23 33 41 43 51 27015 66916 9 16 17 26 33 49 10 12 30 43 46 72 95521 00131 10 15 18 18 43 48 8 11 32 44 47 86 02822 52818 11 19 24 43 46 51 9 36 40 48 50 161 05272 89337 12 13 16 35 35 56 9 19 28 44 55 185 61046 27259 13 9 11 43 45 55 3 19 37 51 53 216 79475 68747 14 9 15 19 34 59 5 10 16 48 57 254 22443 49046 15 23 27 40 49 56 7 39 45 51 53 258 30231 01035 16 8 13 41 45 59 2 10 47 52 55 305 71400 57494 17 1 38 39 39 60 8 25 34 53 57 318 82375 95951

TABLE XII Least n for which a solution to (k. m. n) is known

k

m 2 3 4 5 6 7 8 9 10

1 2 3 4 4 7 8 11 15 23 2 2 2 2 4 7 8 9 12 19 3 3 3 7 8 11 24 4 5 7 10 23 5 5 5 11 16 6 6 27 7 7

except for a solution to (10. 7. 7) given by A. Moessner [35]. Due to computer word length limitations the calculations were not extended to large values of the arguments.

Additional References. A. Gloden gave a parametric solution of (5. 4. 4) in [38], two parametric solutions of (7. 5. 5) in [39], [40], and a parametric solution of (8. 7. 7) in [41]. A. Moessner gave numerical solutions of (5. 2. 4) and (5. 3. 3) in [42]. In [43] Moessner gave three parametric solutions of (6. 4. 4) and parametric solutions of (8. 7. 7) and (9. 10. 10). Two numerical solutions of (7. 4. 5) due to A. Letac are found in [39]. S. Sastry and T. Rai solved (7. 6. 6) parametrically [44].

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 13

458 L. J. LANDER, T. R. PARKIN AND J. L. SELFRIDGE

G. Palam'a [45] gave numerical solutions of (9. 11. 11) and (11. 10. 12). In [46] Moessner and Gloden solved (8. 6. 6) and (8. 6. 7) numerically.

Concluding Remarks. Let N(k,m) be the smallest n for which (k.m.n) is solvable. In Table XII we show the upper bound to N based on the results just presented. Each column is terminated when a solution to (k. m. m) has been found. It appears likely that whenever (k. m. m) is solvable, so is (k. r. r) for any r > m. Some questions are:

(a) Is N(k, m + 1) < N(k, m) < N(k +1, m) always true? (b) Is (k. m. n) always solvable when m + n > k? (c) Is it true that (k. m. n) is never solvable when m + n < k? (d) For which k, m, n such that m + n = k is (k. m. n) solvable? The results presented in this paper tend to support an affirmative answer to (c). Question (d) appears to be especially difficult. The only solvable cases with m + n =k known at present are (4. 2. 2), (5. 1. 4) and (6. 3. 3). In this paper we have made a computational attack on the problem of finding a sum of n kth powers which is also the sum of a smaller number of kth powers. In many of the cases considered, especially for the larger values of k, we have undoubtedly not obtained the best possible results, but the amount of computing needed to do this would seem to be overwhelming. We believe that the main result of this paper is the presentation of results on a family of Diophantine equations which have largely been considered separately in the past. We hope that this presentation offers greater insight into the nature of the function N(k, m) and that future efforts will be directed toward reducing the upper bounds for this function.

Aerospace Corporation Los Angeles, California 90045 Aerospace Corporation Los Angeles, California 90045

Pennsylvania State University University Park, Pennsylvania 16802

1. G. H. HARDY & E. M. WRIGHT, An Introduction to the Theory of Numbers, 3rd ed., Clarendon Press, Oxford, 1954; 4th ed., pp. 332-335, Oxford Univ. Press, London, 1960. MR 16, 673. 2. L. E. DICKSON, History of the Theory of Numbers, Vol. 2, chapters 4-9, 21-24, Publication No. 256, Carnegie Institution of Washington, Washington, D. C., 1920; reprint, Stechert, New York, 1934. 3. L. E. DICKSON, ibid., pp. 165-170. 4. L. E. DICKSON, ibid., p. 252. 5. J. L. SELFRIDGE & B. W. POLLACK, Notices Amer. Math. Soc., no. 11, 1964, p. 97. 6. L. E. DICKSON, loc. cit., pp. 550-554. 7. L. E. DICKSON, loc. cit., pp. 563-565. 8. M. WARD, "Euler's problem on sums of three fourth powers," Duke Math J., v. 15, 1948, pp. 827-837. MR 10,283. 9. R. NORRIE, University of St. Andrews 500th Anniversary Memorial Volume, Edinburgh, 1911, pp. 87-89. 10. J. LEECH, "On A4 + B4 + C4 + D4 = E4," Proc. Cambridge Philos. Soc., v. 54, 1958, pp. 554-555. MR 20 # 2301. 11. S. BRUDNO, "A further example of A4 + B4 + C4 + D4 = E4," Proc. Cambridge Philos. Soc., v. 60, 1964, pp. 1027-1028. MR 29 $#3429. 12. E. FAUQUEMBERGUE, L'intermediaire des Math., v. 5, 1898, p. 33. 13. C. HALDEMAN, Math. Mag., v. 2, 1904, pp. 288-296.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

#### Page 14

A SURVEY OF EQUAL SUMS OF LIKE POWERS 459

14. A. MARTIN, Deux. Congres Internat. Math., 1900, Paris, 1902, pp. 239-248; reprodtuced with additions in Math. Mag., v. 2, 1910, pp. 324-352. 15. L. EULER, Nova Acta Acad. Petrop., 13, ad annos 1795-1796, 1802 (1778), 45; Comm. Arith., II, 281. Cited in L. E. Dickson, loc. cit., pp. 645-646. 16. A. G1ERARDIN, L'intermediaire des Math., v. 24, 1917, p. 51. 17. J. LEECH, "Some solutions of Diophantine equations," Proc. Cambridge Philos. Soc., v. 53, 1957, pp. 778-780. MR 19, 837. 18. L. LANDER & T. PARKIN, "Equal sums of biquadrates," Math. Comp., v. 20, 1966, pp. 450-451.

19. A. GE'RARDIN, Assoc. franC., v. 39, 1910, I, pp. 44-55; same in Sphinx-Oedipe, v. 5, 1910, pp. 180-186; v. 6, 1911, pp. 3-6; v. 8, 1913, p. 119. 20. F. FERRARI, L'intermediaire des Math., v. 20, 1913, pp. 105-106. 21. A. G1ERARDIN, Bull. Soc. Philomathique, (10), v. 3, 1911, p. 236. 22. A. WEREBRUSOW, L'interm6diaire des Math., v. 20, 1913, pp. 105-106. 23. L. LANDER & T. PARKIN, "A counterexample to Euler's conjecture on like powers," Bull. Amer. Math. Soc., v. 72, 1966, p. 173.

24. L. LANDER & T. PARKIN, "A counterexample to Euler's sum of powers conjecture," Math. Comp., v. 21, 1967, pp. 101-103. 25. L. E. DICKSON, loc. cit., p. 648. 26. S. SASTRY, "On sums of powers," J. London Math. Soc., v. 9, 1934, pp. 242-246. 27. A. MARTIN, Bull. Philos. Soc. WVash., v. 10, 1887, p. 107, in Smithsonian Miscel. Coll., v. 33, 1888. 28. K. SUBBA RAO, "On sums of fifth powers," J. London Math. Soc., v. 9, 1934, pp. 170-171. 29. G. XEROUDAKES & A. MOESSNER, "On equal sums of like powers," Proc. Indian Acad. Sci. Sect. A, v. 48, 1958, pp. 245-255. MR 21 #20. 30. A. MARTIN, Quart. J. Math., v. 26, 1893, pp. 225-227. 31. A. MOESSNER, "Einige zahlentheoretische Untersuchungen und diophantische Probleme," Glasnik Mat.-Fiz. Astronom. Drustvo Mat. Fiz. Hrvatske Ser. II, v. 14, 1959, pp. 177-182. M1\R 24 #A2558.

32. K. SUBBA RAO, "On sums of sixth powers," loc. cit., pp. 172-173. 33. A. LETAC, Gazeta Matematica, v. 48, 1942, pp. 68-69. 34. J. 0. PATTERSON, "A note on the Diophantine problem of finding four biquadrates whose sum is a biquadrate," Bull. Amer. Math. Soc., v. 48, 1942, pp. 736-737. MR 4,33. 35. A. MOESSNER, "Einige numerische Identitaten," Proc. Indian Acad. Sci. Sect. A, V. 10, 1939, pp. 296-306. MR 1, 133. 36. H. P. F. SWINNERTON-DYER, "A solution of Al + B5 + C5 = D5 + E5 + F5," Proc. Cambridge Philos. Soc., v. 48, 1952, pp. 516-518. MR 13, 913. 37. L. E. DICKSON, loc. cit., p. 644. 38. A. GLODEN, "Uber mehrgradige Gleichungen," Arch. Math., v. 1, 1949, pp. 482-483. MR 11, 82. 39. A. GLODEN, "Sur la multigrade A1 A2, A3, A4, A5k = B1, B2, B3, B4, B5 (k = 1, 3, 5, 7)," Revista Euclides, v. 8, 1948, pp. 383-384. MVIR 10, 431. 40. A. GLODEN, "Zwei Parameterldsungen einer mehrgradigen Gleichung," Arch. Math., v. 1, 1949, pp. 480-482. MR 11, 82. 41. A. GLODEN, "Parametric solutions of two multi-degreed equalities," Amer. Math. Monthly, v. 55, 1948, pp. 86-88. MR 9, 331.

42. A. MOESSNER, "Alcune richerche di teoria dei numeri e problemi diofantei," Bol. Soc. Mat. Mexicana (5), v. 2, 1948, pp. 36-39. MR 10, 592. 43. A. MOESSNER, "On equal sums of like powers," Math. Student, v. 15, 1947, pp. 83-88. MR 11, 500. 44. S. SASTRY & T. RAI, "On equal sums of like powers," Math. Student, v. 16, 1948, pp. 18-19. MVR 11, 391. 45. G. PALAMA, "Diophantine systems of the type Ei_1 - = ="_b, (k = 1,2, *I*, n, n + 2, n + 4, * * *, n + 2r)," Scripta Math., v. 19, 1953, pp. 132-134. MR 15, 199. 46. A. MOESSNER & A. GLODEN, "Einige Zahlentheoretische Untersuchungen und Resultate," Bull. Sci. Ecole Polytech. de Timisoara, v. 11, 1944, pp. 196-219. MR 9, 9.

This content downloaded from 132.77.150.148 on Sun, 13 Dec 2015 18:40:42 UTC

All use subject to JSTOR Terms and Conditions

---

Source: [Fermat's Library](https://fermatslibrary.com/) — https://fermatslibrary.com/s/a-survey-of-equal-sums-of-like-powers. More annotated papers: https://fermatslibrary.com/llms.txt
